A new study has identified distinct neural and behavioural characteristics in autism that a simple computational principle can explain. This research focuses on the “dynamic range” of neurons, which indicates how gradually or sharply they respond to stimuli. The findings suggest that individuals with autism spectrum disorder exhibit an increased dynamic range in their neuronal responses. As a result, they tend to have more detailed responses that are slower to adapt to changes. This study challenges previous views of autism as merely a “broken cog in the machine” and offers a more nuanced understanding of the computational foundations of autism.
Researchers Dr Yuval Hart and Oded Wertheimer from the Psychology Department and the Edmond and Lily Safra Center for Brain Science (ELSC) at the Hebrew University of Jerusalem have developed a new computational model to explain the neural and behavioural differences associated with autism. This model provides new insights into information processing in the brains of individuals with autism, paving the way for future research and deeper understanding.
Autism is known to present unique neural and behavioural characteristics compared to neurotypical individuals, but the underlying computational mechanisms remain complex and multifaceted. The model proposed by Dr Hart and Wertheimer centres around the concept of “dynamic range” within neuronal populations. Dynamic range refers to signals for which neurons elicit discernible responses. In simpler terms, it reflects how gradually or sharply neurons respond to stimuli – a more gradual response entails an increased dynamic range.
“Our model suggests that autism spectrum disorder is not a `broken cog in the machine`, but rather a spectrum of points on the computational tradeoff line between accurate inference and fast adaptation,” said Dr. Yuval Hart. “This computational trade-off proved to be a fruitful framework for explaining many of the neural and behavioural characteristics seen in autism.”
The researchers found that an increased dynamic range, indicating a gradual response of a neuronal population to changes in input, accounts for neural and behavioural variations in individuals diagnosed with autism across diverse tasks. This gradual response enables more accurate encoding of details but comes with a trade-off: slower adaptation to changes. By contrast, a narrower dynamic range allows for quick, threshold-based reactions, facilitating fast adaptation but potentially at the expense of fine discrimination.
Testing their model across diverse simulations and behavioural tasks, including finger-tapping synchronization, orientation reproduction, and global motion coherence, the researchers demonstrated that increased dynamic range might underlie specific autism r-related behaviours. This variation in response could stem from differences in how individual neurons activate. For instance, increased variability in individual neurons’ half-activation point, where the neuron’s response is half of its maximal value, could lead to a broader dynamic range at the population level, influencing how sensory input is processed and interpreted by the brain.
“We show how heterogeneity in the half-activation points of single neurons can result in a more gradual population response and can thus lead to an increased dynamic range,” explained Oded Wertheimer. “Given the vast literature that maps autism to mutations in genes related to neuronal receptors, the proposed biological mechanism is highly relevant – many sources of heterogeneity in these neuronal features can lead to increased dynamic range. This model provides a new lens for understanding autism spectrum disorder that bridges biological mechanisms at the neuronal level with computational principles.”
Dr. Hart and Wertheimer’s findings also provide insight into why autism spectrum disorder research often yields conflicting results. Differences in the dynamic range within the autism spectrum disorder population may contribute to the variation in findings across studies, highlighting the need for larger participant groups to ensure robust results.
Their model aligns with existing theories that relate autism spectrum disorder to atypical sensory processing, supporting a connection to broader biological and genetic factors. Specific genetic mutations associated with autism spectrum disorder, such as those affecting synaptic regulation, may contribute to this increased dynamic range. These biological factors could lead to a more variable neuronal response, creating the nuanced, analogue-like encoding seen in individuals with autism spectrum disorder individuals.
By exploring this computational trade-off, the study not only introduces a new perspective on autism spectrum disorder but also suggests future directions for study. The researchers propose that examining this dynamic range at different developmental stages or through animal models could further clarify its impact on autism spectrum disorder-related behaviors.
